Digital biomarkers can open a world of possibilities in understanding the nuances of a patient’s behavior and response to treatment. Rather than relying on subjective perceptions of how a patient is responding, digital biomarkers treat patient observation more like an engineering problem. Such precise, sensitive measurements can ultimately influence the future of a patient’s care plan or even the future of a trial. But, just like all AI-powered tools, this potential relies on their ability to work effectively and accurately with the intended patient population – all of which is dependent on the extent to which they embrace diversity.
The development of digital biomarker algorithms is larger than any one company or developer. While we are proud of and confident in our contributions in this field, the true impact of digital biomarkers goes far beyond what we can achieve ourselves. Understanding our responsibility to ensure these tools are developed in the right way, we recognized the opportunity that an open-source platform has to help drive success through diversity. By aggregating diverse perspectives across research and academic communities to interrogate and contribute to the diversity of these novel methodologies, we advance their ability to meet the needs of all patient populations.
Expanding diversity through open communities
We’ve spoken about how diversity is one of our company’s founding principles, and our diligence when building our algorithms on diverse data foundations. This commitment to diversity extends into our need to understand the diverse applications of digital biomarker algorithms and the diverse perspectives the broader community can bring to their evolution. These AI-powered tools and the diversity of the data they are fed need to be considered in light of the patient population and situation they are used in. For example, diversity for a sickle cell population is different than that of cystic fibrosis, and the diversity needed for an AI-powered population screening tool is different from one used to track a patient’s symptoms. Developing these solutions in a proprietary, siloed manner would be a disservice to their potential – no one company has the resources or time to address the vast use cases these solutions can provide value for and achieve generalizability in doing so.
Open-source platforms allow us to bring in as many industry and academic voices as possible to explore, expand and accelerate the diversity of these novel tools. By opening the doors for researchers to import their own data sets, we not only help algorithms learn, but also determine their performance in a variety of unique use cases and patient populations. A forum for unrestricted commentary on the reliability and diversity of these methods enables us to collectively contribute to a growing body of work around the endless ways these tools may be valuable in understanding disease and patient behavior, ultimately guiding their advancement.
Fulfilling the promise of novel assessments
The expansion and validation of these emerging approaches to measuring disease could revolutionize how we draw conclusions about a drug’s impact and its value in real-world settings. But, this is only possible if we are able to build them in the right way with diverse input. Open-source approaches to AI that allow others to explore and contribute to the diversity of algorithms is key to moving digital biomarkers forward. Through deep cross-industry collaboration and building trust in the public domain, we can deliver on the promise of technology to drive equitable, precise care.
